Optimisation of shaft voltage based condition monitoring in generators using a Bayesian approach

This paper presents a framework for the optimisation of shaft voltage based condition monitoring in synchronous generators utilising Bayesian classification. With machines involved in critical processes such as power generation, it is preferable to determine faults well in advance. The proposed system uses shaft voltage signals as an online method for diagnosis of incipient faults in synchronous machines. A Naive Bayes classifier is used in conjunction with frequency spectrum estimation in order to optimise the shaft voltage condition monitoring technique. A Finite Element (FE) model and an experimental machine are used to train, test and validate the fault classification system.